Copper alloy surface drawing quality detection method and device

Abstract

The present invention discloses a copper alloy surface wire drawing quality detection method and device, wherein the copper alloy surface wire drawing quality detection method includes surface cleaning and drying of the copper alloy wire drawing sample to obtain a clean sample; selecting multiple predetermined positions on the surface of the clean sample based on surface feature analysis, performing electrochemical measurements on each predetermined position in turn, and obtaining multiple sets of electrochemical measurement data, wherein the electrochemical measurement includes open circuit potential measurement, electrochemical impedance spectroscopy test and cyclic voltammetry test; performing parameter extraction and mathematical operations on the multiple sets of electrochemical measurement data to obtain an electrochemical characteristic index set characterizing the surface characteristics of the current sample; inputting the electrochemical characteristic index set into a support vector machine classification model pre-trained based on known quality grade sample data, and outputting a classification result representing the quality grade of the current sample. The technical solution of the present invention can improve the reliability of the measurement results.

Description

Technical Field

[0001] The invention relates to the technical field of surface wire drawing quality detection, and in particular to a copper alloy surface wire drawing quality detection method and device. Background Art

[0002] Surface brushing is an important surface treatment process, which is widely used in architectural decoration, electronic equipment housing, high-end furniture and other fields. The brushing process can not only improve the aesthetics of the product, but also enhance the wear resistance and corrosion resistance of the surface. However, the quality of brushing directly affects the appearance and performance of the product, so it is very important to accurately and objectively detect the quality of brushing. A high-quality brushed surface should have a uniform texture, appropriate roughness and good surface integrity, which are directly related to the market competitiveness and service life of the product.

[0003] A thin oxide film will naturally form on the copper surface in the air, and the characteristics of this oxide film are closely related to the quality of wire drawing. A high-quality wire drawing surface will form a uniform and dense oxide layer, providing better protection; while a low-quality wire drawing surface may result in an uneven or loose oxide layer, reducing the corrosion resistance of the product. Traditional wire drawing quality detection methods mainly rely on surface roughness measurement and visual inspection, which are often subjective and difficult to accurately evaluate the characteristics of the oxide layer and the uniformity of the surface microstructure. The measurement results are not comprehensive and reliable enough to reflect the electrochemical characteristics and microstructural changes of the wire drawing surface. Therefore, there is an urgent need for a detection method that can objectively, comprehensively and accurately evaluate the surface drawing quality of copper alloys to meet the strict requirements of modern industrial production for product quality control. Summary of the invention

[0004] The main purpose of the present invention is to solve the technical problem that the measurement results of the existing surface drawing quality detection method are not comprehensive and reliable enough.

[0005] A first aspect of the present invention provides a copper alloy surface drawing quality detection method, the copper alloy surface drawing quality detection method comprising:

[0006] The copper alloy wire drawing samples were surface cleaned and dried to obtain clean samples;

[0007] Based on the surface feature analysis, multiple predetermined positions are selected on the surface of the clean sample, and electrochemical measurements are performed on each predetermined position in turn to obtain multiple sets of electrochemical measurement data, wherein the electrochemical measurements include open circuit potential measurement, electrochemical impedance spectroscopy test and cyclic voltammetry test;

[0008] Performing parameter extraction and mathematical operations on the multiple sets of electrochemical measurement data to obtain a set of electrochemical characteristic indicators characterizing the surface characteristics of the current sample;

[0009] The electrochemical characteristic index set is input into a support vector machine classification model pre-trained based on sample data of known quality grades, and a classification result representing the quality grade of the current sample is output.

[0010] Optionally, the selecting a plurality of predetermined positions on the surface of the clean sample based on the surface feature analysis includes:

[0011] Performing three-dimensional laser scanning on the surface of the clean sample to obtain a microscopic topographic map;

[0012] Performing curvature analysis and shape analysis on the surface of the clean sample according to the microscopic topography map to obtain a surface feature distribution map;

[0013] Performing grid division and spatial clustering based on the surface feature distribution map to obtain a candidate measurement point set;

[0014] Performing Voronoi diagram analysis and spatial autocorrelation analysis on the candidate measurement point set to obtain an optimized measurement point set;

[0015] The points in the optimized measurement point set are determined as predetermined positions to obtain a plurality of predetermined positions.

[0016] Optionally, performing electrochemical measurement on each predetermined position in sequence to obtain multiple sets of electrochemical measurement data includes:

[0017] A movable electrolytic cell is arranged at a current predetermined position, a sample surface exposed in the electrolytic cell is used as a working electrode, a reference electrode and a counter electrode are placed in the electrolytic cell, and an electrolyte is injected into the electrolytic cell to form a local three-electrode measurement system;

[0018] Applying alternating potential and current scans to the working electrode in the local three-electrode measurement system to obtain a working electrode with stable electrochemical activity;

[0019] Sequentially performing open circuit potential measurement, electrochemical impedance spectroscopy test and cyclic voltammetry test on the working electrode with stable electrochemical activity to obtain a set of electrochemical measurement data at the current predetermined position;

[0020] Moving the movable electrolytic cell to the next predetermined position, repeating the above electrolytic cell setting, electrode activation and electrochemical measurement steps until the measurement of all predetermined positions is completed;

[0021] The electrochemical measurement data of all predetermined positions are summarized in position order to obtain multiple groups of electrochemical measurement data including open circuit potential data, impedance spectrum data and voltammetric curve data corresponding to each predetermined position.

[0022] Optionally, the electrochemical impedance spectroscopy test includes:

[0023] Applying multi-frequency AC potential perturbations to the working electrode with stable electrochemical activity to obtain preliminary impedance response data in a wide frequency domain;

[0024] Calculate the frequency derivative of impedance based on the broadband preliminary impedance response data, identify the characteristic frequency interval where impedance changes significantly, and obtain the key frequency range;

[0025] Perform high-resolution frequency scanning within the critical frequency range to obtain local fine impedance response data;

[0026] The broadband preliminary impedance response data and the local fine impedance response data are combined to construct a composite impedance spectrum to obtain impedance spectrum data.

[0027] Optionally, the performing of parameter extraction and mathematical operations on the multiple sets of electrochemical measurement data to obtain a set of electrochemical characteristic indicators characterizing the surface characteristics of the current sample includes:

[0028] Performing statistical analysis on the open circuit potential data of the plurality of groups of electrochemical measurement data, extracting the potential mean and fluctuation range, and obtaining a surface electrochemical stability index;

[0029] Performing equivalent circuit fitting on the impedance spectrum data of the multiple sets of electrochemical measurement data, extracting charge transfer resistance and double layer capacitance values, and obtaining interface electrochemical characteristic indicators;

[0030] Performing peak analysis and integral calculation on the voltammetric curve data of the multiple sets of electrochemical measurement data, extracting redox peak potential, peak current density and reaction charge, and obtaining electrochemical activity indicators;

[0031] Normalizing the surface electrochemical stability index, the interface electrochemical characteristic index and the electrochemical activity index to obtain a standardized electrochemical characteristic index;

[0032] The standardized electrochemical characteristic indexes are subjected to principal component analysis to extract main characteristic vectors, thereby obtaining a set of electrochemical characteristic indexes after dimensionality reduction.

[0033] Optionally, the principal component analysis is performed on the standardized electrochemical characteristic index to extract the main characteristic vectors to obtain a set of electrochemical characteristic indicators after dimension reduction, including:

[0034] Constructing a covariance matrix of the standardized electrochemical characteristic index, performing eigenvalue decomposition, and obtaining eigenvalues ​​and corresponding eigenvectors;

[0035] Sort the eigenvectors according to the eigenvalues, calculate the cumulative contribution rate, and determine the number of principal components to be retained;

[0036] Performing a projection transformation on the standardized electrochemical characteristic index using the determined number of principal components to obtain a principal component score after preliminary dimensionality reduction;

[0037] Performing orthogonal rotation on the principal component scores after the preliminary dimensionality reduction to obtain rotated principal components;

[0038] The rotated principal components are used as a set of electrochemical characteristic indicators after dimensionality reduction.

[0039] Optionally, the training process of the support vector machine classification model includes:

[0040] Conduct electrochemical measurements and data processing on copper alloy wire drawing samples with known quality grades to obtain a training data set with quality grade labels.

[0041] Performing feature normalization and dimensionality reduction processing on the training data set to obtain a preprocessed feature matrix;

[0042] Constructing a kernel matrix according to the preprocessed feature matrix, performing grid search on different kernel functions and parameter combinations to obtain an optimal kernel function configuration;

[0043] Using the optimal kernel function configuration and the preprocessed feature matrix, the dual problem of the support vector machine is solved by a sequential minimum optimization algorithm to obtain a support vector and a decision hyperplane;

[0044] The support vector and the decision hyperplane are probability calibrated to obtain a support vector machine classification model with probability output capability.

[0045] Optionally, the electrochemical characteristic index set is input into a support vector machine classification model pre-trained based on sample data of known quality grades, and a classification result indicating the quality grade of the current sample is output, including:

[0046] Normalizing and reducing the dimension of the electrochemical characteristic index set to obtain a standardized feature vector consistent with the training data format;

[0047] According to the standardized feature vector and the kernel function of the support vector machine classification model, a kernel matrix of the current sample and the support vector is calculated to obtain a decision function value;

[0048] Applying a probability calibration function to the decision function value to obtain a probability distribution of the sample belonging to each quality grade;

[0049] Performing multi-level threshold judgment according to the probability distribution to obtain the primary quality grade and secondary quality grade of the sample;

[0050] The primary quality grade, the secondary quality grade and their corresponding probability values ​​are combined to obtain a classification result representing the quality grade of the current sample.

[0051] Optionally, performing multi-level threshold judgment according to the probability distribution to obtain the primary quality grade and the secondary quality grade of the sample includes:

[0052] Performing entropy analysis on the probability distribution to obtain an uncertainty index of the sample quality grade distribution;

[0053] A dynamic threshold strategy is selected according to the uncertainty index, a single threshold method is used for low uncertainty samples, and a multiple threshold method is used for high uncertainty samples to obtain a threshold judgment scheme;

[0054] Applying the threshold judgment scheme to the probability distribution to obtain a preliminary quality grade judgment result;

[0055] Performing a neighborhood consistency check on the preliminary quality grade judgment result, and performing local smoothing according to the quality grade distribution of adjacent measurement positions to obtain a spatially consistent quality grade distribution;

[0056] The grade with the highest probability is extracted from the spatially consistent quality grade distribution as the primary quality grade, and the grade with the second highest probability and exceeding a preset threshold is extracted as the secondary quality grade, to obtain the primary quality grade and secondary quality grade of the sample.

[0057] A second aspect of the present invention provides a copper alloy surface drawing quality detection device, comprising:

[0058] The sample pretreatment unit is used to clean and dry the surface of the copper alloy wire drawing sample to obtain a clean sample;

[0059] An electrochemical measurement module, for selecting a plurality of predetermined positions on the surface of the clean sample based on surface feature analysis, performing electrochemical measurement on each predetermined position in turn, and obtaining a plurality of sets of electrochemical measurement data, wherein the electrochemical measurement includes open circuit potential measurement, electrochemical impedance spectroscopy test and cyclic voltammetry test;

[0060] A data processing unit, used for performing parameter extraction and mathematical operations on the multiple sets of electrochemical measurement data to obtain a set of electrochemical characteristic indicators characterizing the surface characteristics of the current sample;

[0061] The quality classification module is used to input the electrochemical characteristic index set into a support vector machine classification model pre-trained based on sample data of known quality grades, and output a classification result representing the quality grade of the current sample.

[0062] The copper alloy surface drawing quality detection method proposed in the present invention effectively solves the limitations of traditional detection methods and provides a comprehensive, objective and accurate quality assessment scheme. First, the method adopts a multi-point measurement strategy based on surface feature analysis to overcome the limitations of single-point measurement of traditional methods. By selecting multiple predetermined positions on the sample surface for measurement, it is not only possible to comprehensively evaluate the quality consistency of the entire drawing surface, but also to capture local quality changes. Secondly, the method uses three electrochemical measurement techniques: open circuit potential measurement, electrochemical impedance spectroscopy test and cyclic voltammetry test to obtain rich surface electrochemical information. Open circuit potential measurement reflects the thermodynamic state of the surface, electrochemical impedance spectroscopy test provides information about the structure of the oxide layer and the charge transfer process, and cyclic voltammetry test reveals the redox behavior of the surface. This multi-dimensional electrochemical characterization can not only evaluate the surface morphology, but also deeply analyze the characteristics of the oxide layer and the surface electrochemical activity, thereby providing a more comprehensive and in-depth quality assessment than traditional methods.

[0063] In addition, this method obtains a set of electrochemical characteristic indicators that characterize the surface characteristics of the sample by performing parameter extraction and mathematical operations on multiple sets of electrochemical measurement data. This process not only eliminates the influence of human subjective factors, but also captures tiny changes in surface characteristics. Finally, the support vector machine classification model is used to determine the quality grade, converting complex multidimensional data into clear quality grade results. This machine learning-based classification method has the characteristics of strong adaptability and good scalability, and can be flexibly adjusted according to different quality standards and application requirements. At the same time, since the model is trained based on a large amount of known quality grade sample data, it has a strong generalization ability and can accurately identify various types and degrees of wire drawing quality problems. In general, this method solves the local problem of traditional methods through a multi-point measurement strategy, provides more comprehensive surface information through multi-dimensional electrochemical characterization, and improves the accuracy and objectivity of detection through advanced data processing and machine learning technology, thereby realizing a comprehensive, accurate and objective evaluation of the surface wire drawing quality of copper alloys, providing strong technical support for product quality control. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0065] Figure 1 A schematic diagram of an embodiment of a method for detecting the surface wire drawing quality of a copper alloy in an embodiment of the present invention;

[0066] Figure 2 Schematic diagram of an embodiment of a copper alloy surface drawing quality detection device in an embodiment of the present invention.

[0067] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0068] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0069] It should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back...), the directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0070] In addition, the descriptions of "first", "second", etc. in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, "and / or" in the full text includes three solutions. Taking A and / or B as an example, it includes technical solution A, technical solution B, and technical solution that satisfies both A and B. In addition, the technical solutions between the various embodiments can be combined with each other, which must be based on the ability of ordinary technicians in the field to implement. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0071] An embodiment of the present application provides a method for detecting the surface wire drawing quality of a copper alloy. Figure 1 A flow chart of a copper alloy surface drawing quality detection method provided in one embodiment of the present application.

[0072] See also Figure 1 , the copper alloy wire drawing sample was surface cleaned and dried to obtain a clean sample.

[0073] Specifically, first, the copper alloy wire drawing sample is placed in an ultrasonic cleaner and preliminarily cleaned using anhydrous ethanol as the cleaning solvent. Ultrasonic cleaning uses the cavitation effect generated by sound waves in the liquid to effectively remove oil, dust and other organic contaminants on the surface of the sample. This step usually lasts 10-15 minutes, during which time it can be observed that the cleaning liquid gradually becomes turbid, indicating that the contaminants have been successfully removed. Subsequently, the sample is rinsed multiple times with deionized water to ensure that all ethanol and dissolved contaminants are completely removed. The use of deionized water can prevent minerals in the water from leaving residues on the surface of the sample.

[0074] Optionally, to remove any stubborn contaminants that may be present, a mild surfactant solution can be used for further cleaning without affecting the surface oxide layer of the sample. For example, gently wiping the sample surface with a 0.1% neutral detergent solution can effectively remove residual organic contaminants and particulates without affecting the surface oxide layer. The use of a neutral detergent ensures that the cleaning process does not change the electrochemical properties of the sample surface, which is critical for subsequent oxide layer analysis. After cleaning, rinse the sample thoroughly again with deionized water to remove all detergent residues.

[0075] After the cleaning process is completed, the drying process begins. First, the sample surface is purged with high-purity nitrogen to quickly remove most of the moisture. The inert properties of nitrogen prevent further oxidation reactions on the sample surface during the drying process. The sample is then placed in a vacuum drying oven and vacuum dried at room temperature (about 25°C) for 30 minutes. The vacuum environment accelerates the evaporation of water while preventing oxygen and other contaminants in the air from coming into contact with the sample. This room temperature vacuum drying method can effectively prevent microstructural changes on the sample surface due to temperature changes, and ensure the original surface characteristics of the sample, including the integrity of the oxide layer.

[0076] Each step of the entire cleaning and drying process is carefully designed to ensure the best cleaning effect while retaining the original state of the surface oxide layer. For example, in a typical treatment, a 10cm×10cm copper alloy wire drawing sample can be treated with the above treatment, and its surface contact angle can be reduced from the initial 70° to about 40°, indicating that the surface cleanliness is significantly improved while still retaining a certain degree of hydrophobicity. In addition, through X-ray photoelectron spectroscopy (XPS) analysis, it can be found that the carbon content on the surface of the sample is significantly reduced after treatment, while the intensity and shape of the oxide peak remain basically unchanged, which proves that the cleaning process effectively removes contaminants while retaining the surface oxide layer.

[0077] This optimized cleaning and drying process has several advantages. First, it can completely remove various types of contaminants, including organic and inorganic particles, without affecting the surface oxide layer, ensuring the accuracy of subsequent electrochemical measurements. Second, the use of mild cleaning conditions and room temperature drying can maximize the preservation of the original surface morphology, microstructure and oxide layer state of the sample, avoiding the introduction of additional surface changes. Furthermore, the vacuum drying process effectively prevents further oxidation of the sample and ensures the long-term stability of the sample. Finally, this standardized processing flow has good repeatability, which is conducive to comparative analysis between different batches of samples.

[0078] Through this series of fine cleaning and drying treatments, various contaminants on the surface of the copper alloy wire drawing samples are effectively removed, while maintaining the original surface microstructure and oxide layer. The final clean sample has a uniform and stable surface state, laying a solid foundation for subsequent electrochemical measurements and oxide layer analysis. This high-quality sample pretreatment not only improves the accuracy and reliability of the measurement results, but also ensures the scientificity and repeatability of the entire quality inspection process.

[0079] Please continue reading Figure 1 Based on the surface feature analysis, multiple predetermined positions are selected on the surface of the clean sample, and electrochemical measurements are performed on each predetermined position in turn to obtain multiple sets of electrochemical measurement data, wherein the electrochemical measurement includes open circuit potential measurement, electrochemical impedance spectroscopy test and cyclic voltammetry test.

[0080] In one embodiment of the present invention, the method of selecting multiple predetermined positions on the surface of the clean sample based on surface feature analysis includes: performing three-dimensional laser scanning on the surface of the clean sample to obtain a microscopic topographic map; performing curvature analysis and shape analysis on the surface of the clean sample according to the microscopic topographic map to obtain a surface feature distribution map; performing grid division and spatial clustering based on the surface feature distribution map to obtain a candidate measurement point set; performing Voronoi diagram analysis and spatial autocorrelation analysis on the candidate measurement point set to obtain an optimized measurement point set; and determining points in the optimized measurement point set as predetermined positions to obtain multiple predetermined positions.

[0081] Specifically, a confocal laser scanning microscope is used to scan the entire sample surface with a resolution of micrometers. For example, for a 10cm×10cm sample, the scanning step size can be set to 10μm, thereby obtaining 10 million data points to form a highly accurate microscopic topography map. This high-resolution 3D scan can not only capture the fine texture left by the wire drawing process, but also reflect the tiny defects and unevenness of the surface.

[0082] Next, the sample surface is subjected to in-depth curvature analysis and shape analysis based on the obtained microscopic topography. Curvature analysis uses two indicators, Gaussian curvature and mean curvature. By calculating the local curvature value of each point, the convex, concave and flat areas of the surface can be identified. At the same time, shape analysis focuses on the directionality, periodicity and complexity of the surface texture. The combination of these two analysis methods can fully capture the geometric characteristics of the surface. By visualizing the analysis results, a color-coded surface feature distribution map is generated, in which different colors represent different curvature values ​​and shape features. For example, red represents high curvature areas, blue represents low curvature areas, and the direction of the texture can be represented by arrows. This intuitive surface feature distribution map provides an important basis for subsequent analysis.

[0083] Based on the generated surface feature distribution map, the next step is to perform grid division and spatial clustering to determine the candidate measurement point set. First, the entire sample surface is divided into a uniform grid. The grid size can be determined according to the sample size and the desired measurement accuracy, usually 1mm×1mm. Then, the features within each grid cell are statistically analyzed to calculate parameters such as the average curvature and the main texture direction. Using these parameters, clustering algorithms such as K-means are used to spatially cluster the grids, and areas with similar features are grouped together. This method can effectively identify typical and special areas of the surface. For example, on a sample with a clear brushed texture, strip-like clusters parallel to the brushing direction may be formed.

[0084] In order to further optimize the distribution of measurement points, Voronoi diagram analysis and spatial autocorrelation analysis are performed on the candidate measurement point set. Voronoi diagram analysis is a spatial partitioning method that divides the plane into several regions. The distance from the point in each region to the seed point corresponding to the region (in this case, the candidate measurement point) is smaller than the distance to any other seed point. In practical applications, the Voronoi diagram is first constructed with the initial candidate point as the seed point, and then the area of ​​each Voronoi polygon is calculated. If the area of ​​some polygons is significantly larger or smaller than the average, it means that the sampling points in the corresponding area are unevenly distributed. By adjusting the position of the seed point or adding or deleting the seed point, the area of ​​all Voronoi polygons is made as close as possible, thereby ensuring that the selected points can evenly cover the entire sample surface and avoid overly concentrated or sparse areas.

[0085] Spatial autocorrelation analysis is used to evaluate the spatial distribution pattern of features. By calculating statistics such as the Moran's I index, areas with significant spatial correlation can be identified. The Moran's I index is an important statistical tool used to measure the similarity of attribute values ​​between spatial units, and its value range is usually between -1 and 1. When the Moran's I index is close to 1, it means that observation points with similar values ​​tend to cluster in space; when it is close to -1, it means that observation points with different values ​​are adjacent in space; and when it is close to 0, it means that the observation values ​​are randomly distributed in space. In the surface analysis of copper alloy wire drawing samples, the Moran's I index can help identify the spatial distribution pattern of surface features. For example, if the calculated Moran's I index is 0.75, it indicates that surface features (such as roughness or texture depth) show a strong positive correlation in space, that is, similar feature values ​​tend to cluster together. This information is very valuable for understanding the uniformity of the wire drawing process and identifying possible quality problem areas.

[0086] The combination of Voronoi diagram analysis and Moran's I index analysis ensures the spatial uniformity of the measurement points while taking into account the spatial distribution of surface features. This approach has multiple advantages: First, the Moran's I index provides a quantitative description of the spatial distribution of surface features, which helps to identify areas where quality issues may exist. Second, the Voronoi diagram analysis ensures the spatial uniformity of the measurement points, avoiding oversampling or undersampling in certain areas. This uniform distribution not only improves the representativeness of the sampling, but also optimizes the efficiency of subsequent electrochemical measurements. Furthermore, the combination of these two methods takes into account the spatial correlation of surface features and the spatial distribution of sampling points, providing a scientific basis for selecting the most representative measurement points.

[0087] Finally, the points in the optimized measurement point set are determined as predetermined positions to obtain multiple predetermined positions. These predetermined positions include both typical areas of the surface and special areas, which can fully reflect the characteristics of the sample surface. Usually, for a 10cm×10cm sample, the number of predetermined positions finally selected may be between 50 and 100, depending on the surface complexity and the required measurement accuracy. For example, on a 100mm×100mm copper alloy drawing sample, 80 predetermined positions may be finally selected. Most of these positions are evenly distributed over the entire surface, reflecting the overall characteristics of the sample; at the same time, in areas where the Moran's I index shows high aggregation, several additional measurement points may be added to better characterize these special areas.

[0088] This method of selecting predetermined positions based on high-precision three-dimensional scanning and complex data analysis has several significant advantages. First, it can objectively and comprehensively capture the microscopic features of the sample surface, avoiding the subjectivity and bias that may be caused by manual point selection. Secondly, by comprehensively considering the geometric characteristics and spatial distribution of the surface, the selected predetermined positions are highly representative and can effectively reflect the surface quality of the entire sample. Furthermore, this method has a high degree of automation and can quickly process a large amount of data, making it suitable for quality control in mass production. Finally, since the point selection process takes into account a variety of factors, including curvature, shape, spatial distribution, etc., the obtained set of predetermined positions can provide rich and diverse information for subsequent electrochemical measurements, which is conducive to a comprehensive assessment of wire drawing quality. In general, this method of selecting predetermined positions lays a solid foundation for subsequent electrochemical measurements and significantly improves the scientificity and reliability of the entire quality inspection process. This strategy not only ensures the comprehensiveness of the measurement, but also improves the detection sensitivity of potential quality problem areas.

[0089] In one embodiment of the present invention, the electrochemical measurement is performed on each predetermined position in sequence to obtain multiple sets of electrochemical measurement data, including: setting a movable electrolytic cell at the current predetermined position, using the sample surface exposed in the electrolytic cell as a working electrode, placing a reference electrode and a counter electrode in the electrolytic cell, and injecting an electrolyte into the electrolytic cell to form a local three-electrode measurement system; applying alternating potential and current scanning to the working electrode in the local three-electrode measurement system to obtain a working electrode with stable electrochemical activity; sequentially performing open circuit potential measurement, electrochemical impedance spectrum test and cyclic voltammetry test on the working electrode with stable electrochemical activity to obtain a set of electrochemical measurement data for the current predetermined position; moving the movable electrolytic cell to the next predetermined position, repeating the above steps of electrolytic cell setting, electrode activation and electrochemical measurement until the measurement of all predetermined positions is completed; summarizing the electrochemical measurement data of all predetermined positions in positional order to obtain multiple sets of electrochemical measurement data including open circuit potential data, impedance spectrum data and voltammetric curve data corresponding to each predetermined position.

[0090] Specifically, the movable electrolytic cell allows local measurements at different locations on the sample surface without immersing the entire sample. This electrolytic cell can be formed by a ring-shaped structure (such as cylindrical glass). Taking cylindrical glass as an example, there is a small hole at the bottom of the cylindrical glass with a diameter of about 5 mm, which just covers the predetermined measurement position. Through this small hole, a small area of ​​the sample surface is exposed to the electrolyte and becomes a working electrode. Furthermore, a flexible sealing ring can be installed on the opening periphery at the bottom of the cylindrical glass. The sealing ring is made of a chemically inert elastic material. The cylindrical glass is placed on the clean sample surface. By applying uniform pressure to the cylindrical glass, the sealing ring is tightly fitted to the sample surface, and a sealed cavity can be formed.

[0091] In the electrolytic cell, in addition to the sample surface as the working electrode, a reference electrode and a counter electrode need to be placed. Saturated calomel electrode (SCE) is usually used as the reference electrode because it provides a stable reference potential. Platinum wire or graphite rod is usually used as the counter electrode because these materials have good conductivity and chemical stability. These three electrodes constitute the classic three-electrode measurement system. The advantage of this system is that it can accurately measure the potential of the working electrode without being affected by the solution resistance and the counter electrode reaction.

[0092] Next, the prepared electrolyte is injected into the electrolytic cell. The choice of electrolyte is critical because it directly affects the sensitivity and repeatability of the measurement. For the measurement of copper alloy surfaces, the commonly used electrolyte is 0.1M Solution (meaning that 1 liter of solution contains 0.1 mole of ), this solution has good conductivity and will not cause serious corrosion to the copper surface. After injecting the electrolyte, the entire local three-electrode measurement system is completed.

[0093] Before starting the formal measurement, the working electrode (i.e., the sample surface) needs to be activated. This step is achieved by applying alternating potential and current scans to the working electrode. Specifically, a rapid potential cycle can be performed within the range of ±100mV (vs.SCE), or a short constant current pulse can be applied. The purpose of this process is to remove possible contaminants or passivation layers on the surface so that the electrode surface reaches a stable state of electrochemical activity. This step is critical to obtaining reliable and repeatable measurements because it ensures that each measurement is performed under the same initial conditions.

[0094] After activation is complete, a series of electrochemical measurements can be started. The first is the open circuit potential (OCP) measurement. This test does not apply external current and only measures the equilibrium potential of the electrode in the electrolyte. The OCP value reflects the thermodynamic state of the material surface and is very useful for evaluating the corrosion tendency of the surface. Next is the electrochemical impedance spectroscopy (EIS) test, which measures the impedance response of the system by applying a small amplitude AC voltage perturbation at different frequencies. EIS data can provide rich information about the electrode / electrolyte interface, including double layer capacitance, charge transfer resistance, etc. The last is the cyclic voltammetry (CV) test, which measures the corresponding current response by scanning the potential within a certain potential range. The CV curve can reflect the redox reaction occurring on the electrode surface and is very valuable for studying the electrochemical activity of the surface.

[0095] The combination of these three test methods provides comprehensive electrochemical information, making the analysis of surface characteristics more in-depth and comprehensive. For example, in a typical measurement, OCP may be displayed as -0.2Vvs.SCE, indicating that the surface is in a relatively active state. EIS test may show a semicircle in the high frequency area and a straight line in the low frequency area, corresponding to the charge transfer process and diffusion process, respectively. CV curves may show obvious redox peaks, reflecting the presence of surface active groups.

[0096] After completing the measurement at the current position, move the movable electrolytic cell to the next predetermined position (the electrolyte in the electrolytic cell needs to be taken out first). This process requires careful operation to ensure that the sample surface is not damaged or the electrode is contaminated. At the new position, repeat the above steps of electrolytic cell setup, electrode activation and electrochemical measurement. Although this point-by-point measurement method is time-consuming, it can provide highly localized information and help identify spatial changes in surface properties.

[0097] When all predetermined positions have been measured, the electrochemical measurement data obtained are summarized in position order. This data set contains the open circuit potential data, impedance spectrum data, and voltammetric curve data corresponding to each predetermined position. This multi-dimensional, multi-position data set not only reflects the overall characteristics of the sample surface, but also reveals local changes and inhomogeneities. For example, by comparing the OCP values ​​at different positions, changes in surface activity can be identified; by comparing EIS data, changes in oxide layer thickness can be found; and by analyzing CV curves, differences in electrochemical activity in different regions can be determined.

[0098] This local electrochemical measurement method based on a movable electrolytic cell has several significant advantages. First, it can provide highly localized information, which helps to identify small changes and inhomogeneities in surface properties. Second, by measuring at multiple locations, spatial distribution information of surface properties can be obtained, which is crucial for evaluating the uniformity of wire drawing quality. Furthermore, this method is non-destructive, and the sample can still be used for other analyses or practical applications after the measurement is completed. Finally, this method is highly flexible, and the measurement position and measurement parameters can be adjusted as needed, which is suitable for a variety of different research purposes. In general, this electrochemical measurement method provides a rich and accurate data basis for the evaluation of copper alloy surface drawing quality, significantly improving the accuracy and reliability of quality control.

[0099] In one embodiment of the present invention, the electrochemical impedance spectroscopy test includes: applying a multi-frequency AC potential perturbation to the working electrode with stable electrochemical activity to obtain preliminary impedance response data in a wide frequency domain; calculating the frequency derivative of impedance based on the preliminary impedance response data in the wide frequency domain, identifying the characteristic frequency interval in which the impedance changes significantly, and obtaining a key frequency range; performing a high-resolution frequency scan within the key frequency range to obtain local fine impedance response data; and combining the preliminary impedance response data in the wide frequency domain and the local fine impedance response data to construct a composite impedance spectrum to obtain impedance spectrum data.

[0100] Specifically, the process begins with applying a series of small AC potential perturbations of different frequencies to the working electrode, usually starting from high frequency (such as 100kHz) and gradually decreasing to low frequency (such as 0.01Hz). At each frequency point, the system records the current response of the electrode and calculates the corresponding impedance value through Fourier transform. This wide frequency domain scan can fully reflect the electrochemical properties of the electrode / electrolyte interface, including charge transfer, double layer capacitance, diffusion process, etc.

[0101] For example, in a typical measurement, 10 points may be selected for each frequency level, and a total of 60 frequency points are measured from 100kHz to 0.01Hz. The preliminary impedance response data obtained in this way forms a wide-frequency domain impedance spectrum, which lays the foundation for subsequent in-depth analysis. The advantage of this wide-frequency domain scan is that it can capture electrochemical processes occurring at different time scales, from fast electron transfer to slow ion diffusion, which can be reflected in the spectrum.

[0102] After obtaining the preliminary impedance response data in the wide frequency domain, the next step is to calculate the frequency derivative of the impedance to identify the characteristic frequency intervals where the impedance changes significantly. The core idea of ​​this step is to find the areas where the impedance changes most dramatically with frequency, because these areas often contain the most important electrochemical information. Specifically, the derivatives of the real and imaginary parts of the impedance with respect to frequency can be calculated, or the derivative of the impedance modulus with respect to frequency can be calculated. By analyzing the peak values ​​and changing trends of these derivative values, several key frequency ranges can be determined.

[0103] For example, suppose that when analyzing a copper alloy wire drawing sample, a significant peak is found in the frequency derivative of the impedance between 1kHz and 100Hz, which may mean that an important charge transfer process occurs in this frequency range. At the same time, a large impedance change is also observed between 1Hz and 0.1Hz, which may be related to the diffusion process. Based on these observations, 1kHz to 100Hz and 1Hz to 0.1Hz can be identified as two key frequency ranges.

[0104] After determining the critical frequency ranges, the next step is to perform a high-resolution frequency scan within these ranges. The purpose of this step is to obtain more detailed impedance response data to more accurately characterize the key electrochemical processes. In the above example, an additional 50 measurement points may be added between 1kHz and 100Hz, and 30 measurement points may be added between 1Hz and 0.1Hz. This fine scan can provide more data points, which is helpful for subsequent data fitting and parameter extraction.

[0105] An important advantage of high-resolution scanning is that it can reveal details that might be missed with conventional broadband scanning. For example, when studying a passive film on a copper alloy surface, a fine scan might reveal multiple time constants, reflecting the multi-layer structure of the passive film. This information is critical in evaluating the protective properties of a brushed surface.

[0106] The final step is to combine the broadband preliminary impedance response data with the local fine impedance response data to construct a composite impedance spectrum. This process requires careful data processing to ensure that the data in different frequency ranges can transition smoothly. Typically, data interpolation and smoothing techniques can be used to fuse the two sets of data. The result is a comprehensive impedance spectrum that contains both broadband global information and the fine structure of key areas.

[0107] The advantage of this composite impedance spectrum is that it retains the comprehensiveness of the wide frequency domain scan and contains high-precision information in the key frequency range. For example, when evaluating the uniformity of the copper alloy drawing surface, the wide frequency domain data can provide the overall impedance characteristics, while the fine scanning data can accurately reflect the local electrochemical activity differences.

[0108] In general, this method of combining wide-frequency domain scanning and local fine scanning greatly improves the information content and accuracy of electrochemical impedance spectroscopy testing. It can not only comprehensively characterize the electrochemical properties of the copper alloy wire drawing surface, but also accurately capture subtle changes in key frequency ranges. This is of great significance for a deep understanding of the impact of the wire drawing process on the surface microstructure, as well as for evaluating the uniformity and stability of the wire drawing quality. The high-quality impedance spectroscopy data obtained by this method provides a solid foundation for subsequent data analysis and quality assessment, and helps to improve the accuracy and reliability of copper alloy surface wire drawing quality detection.

[0109] Please continue reading Figure 1 , performing parameter extraction and mathematical operations on the multiple sets of electrochemical measurement data to obtain a set of electrochemical characteristic indicators characterizing the surface characteristics of the current sample;

[0110] In one embodiment of the present invention, the parameter extraction and mathematical operation of the multiple sets of electrochemical measurement data to obtain a set of electrochemical characteristic indicators characterizing the surface characteristics of the current sample include: statistical analysis of the open circuit potential data of the multiple sets of electrochemical measurement data, extraction of the potential mean and fluctuation range, and acquisition of a surface electrochemical stability indicator; equivalent circuit fitting of the impedance spectrum data of the multiple sets of electrochemical measurement data, extraction of charge transfer resistance and double layer capacitance, and acquisition of interface electrochemical characteristic indicators; peak analysis and integral calculation of the voltammetric curve data of the multiple sets of electrochemical measurement data, extraction of redox peak potential, peak current density and reaction charge, and acquisition of electrochemical activity indicators; normalization of the surface electrochemical stability indicator, interface electrochemical characteristic indicator and electrochemical activity indicator to obtain a standardized electrochemical characteristic indicator; principal component analysis of the standardized electrochemical characteristic indicator to extract the main eigenvectors and obtain a reduced-dimensional electrochemical characteristic indicator set.

[0111] Specifically, first, the open circuit potential data is statistically analyzed, a step aimed at evaluating the surface electrochemical stability. Specifically, the mean open circuit potential of all measurement points is calculated, which reflects the average electrochemical state of the entire surface. At the same time, the standard deviation or range of the open circuit potential is calculated as a measure of the fluctuation range. These two parameters together constitute an indicator of surface electrochemical stability. For example, if the mean open circuit potential of a sample is -0.2V (relative to the reference electrode) and the standard deviation is 0.01V, this indicates that the surface has relatively uniform and stable electrochemical properties. In contrast, if the standard deviation reaches 0.05V, it means that there is significant inhomogeneity or local activity differences on the surface.

[0112] Next, the impedance spectrum data is fitted with an equivalent circuit, which is an important method to obtain indicators of the electrochemical properties of the interface. Commonly used equivalent circuit models include Randles circuits or their variants, which can simulate the electrochemical behavior of the electrode / electrolyte interface. Through fitting, key electrochemical parameters such as charge transfer resistance (Rct) and double layer capacitance (Cdl) can be extracted. The charge transfer resistance reflects the ease of electron transfer at the interface, while the double layer capacitance is related to the interface adsorption and surface area. For example, if the Rct value obtained by fitting is low (such as 100Ω·cm²) and the Cdl value is high (such as 50μF / cm²), this may indicate that the surface has high electrochemical activity and a large effective surface area, which is usually a feature of a high-quality drawing surface.

[0113] Subsequently, the voltammetric data is subjected to peak analysis and integral calculation to obtain electrochemical activity indicators. This step involves identifying the location (peak potential) and intensity (peak current density) of the redox peaks, as well as calculating the reaction charge by integration. Together, these parameters reflect the electrochemical activity and reaction kinetics of the surface. For example, when analyzing the surface of a copper alloy, a clear oxidation peak may be observed at approximately 0.2 V (vs. Ag / AgCl), with a peak current density of 2 mA / cm² and an integrated reaction charge of 500 μC / cm². These values ​​provide important information about the number of surface active groups and reaction rates.

[0114] In order to make indicators of different dimensions and orders of magnitude comparable, it is necessary to normalize the surface electrochemical stability indicators, interface electrochemical characteristics indicators and electrochemical activity indicators. Commonly used normalization methods include minimum-maximum normalization or Z-score normalization. This step ensures that different indicators have the same weight in subsequent analysis. For example, the open circuit potential of -0.2V is converted to a normalized value of 0.7, and the Rct value of 100Ω·cm² is converted to a normalized value of 0.3. This treatment allows indicators of different properties to be compared and comprehensively analyzed on the same scale.

[0115] Finally, the standardized electrochemical characteristic indicators are subjected to principal component analysis (PCA), which is a powerful dimensionality reduction and feature extraction technique. PCA can project multidimensional data into a low-dimensional space consisting of several main eigenvectors while retaining the main variation information of the data. In practical applications, the first few principal components that can explain 80% or 90% of the total variance are usually selected. For example, if the original data contains 20 features, PCA may reduce it to 5 principal components, which contain 90% of the information of the original data. This dimensionality reduction not only greatly reduces the complexity of the data, but also helps to remove noise and redundant information and highlight key features.

[0116] This multi-step data processing and analysis method has multiple advantages. First, it comprehensively evaluates surface properties from different perspectives, including stability, interface properties, and activity, providing rich information about the material surface. Second, through standardized processing, different types of data can be comprehensively compared, which enhances the objectivity and comprehensiveness of the analysis. Furthermore, the application of principal component analysis not only reduces the complexity of the data, but also helps to identify the most representative features, providing efficient input for subsequent quality classification. In addition, this method is highly flexible and can adjust analysis parameters and focus according to specific research needs. It is suitable for various types of copper alloy surface drawing quality assessment scenarios.

[0117] In general, this comprehensive electrochemical data analysis method provides a scientific and comprehensive basis for the quantitative evaluation of copper alloy surface drawing quality. It can not only capture the overall characteristics of the surface, but also reflect local changes and subtle differences, greatly improving the accuracy and reliability of quality detection. The reduced-dimensional electrochemical feature index set obtained by this method provides high-quality input data for subsequent machine learning classification models, which helps to achieve automated and intelligent drawing quality evaluation.

[0118] In one embodiment of the present invention, the principal component analysis is performed on the standardized electrochemical characteristic indicators to extract the main eigenvectors to obtain the electrochemical characteristic indicator set after dimensionality reduction, including: constructing the covariance matrix of the standardized electrochemical characteristic indicators, performing eigenvalue decomposition, and obtaining eigenvalues ​​and corresponding eigenvectors; sorting the eigenvectors according to the size of the eigenvalues, calculating the cumulative contribution rate, and determining the number of principal components to be retained; using the determined number of principal components to perform a projection transformation on the standardized electrochemical characteristic indicators to obtain the principal component scores after preliminary dimensionality reduction; performing an orthogonal rotation on the principal component scores after preliminary dimensionality reduction to obtain the rotated principal components; and using the rotated principal components as the electrochemical characteristic indicator set after dimensionality reduction.

[0119] Specifically, in the data analysis process of copper alloy surface drawing quality inspection, principal component analysis (PCA) is used to reduce the data dimension and extract the most important features. The process begins with the construction of the covariance matrix of the standardized electrochemical characteristic indicators. Assuming that there are n samples and each sample has m electrochemical characteristic indicators, then the data first need to be organized into an n×m matrix. Calculate the covariance matrix of this matrix to obtain an m×m symmetric matrix that reflects the correlation between different features. For example, in a data set containing 10 features such as open circuit potential, charge transfer resistance, and double layer capacitance, the covariance matrix will be a 10×10 matrix, where the diagonal elements represent the variance of each feature and the off-diagonal elements represent the covariance between the features.

[0120] Next, the covariance matrix is ​​subjected to eigenvalue decomposition, which will yield m eigenvalues ​​and their corresponding eigenvectors. The eigenvalues ​​represent the variance contribution of each principal component, while the eigenvectors define the direction of the principal component. In practice, this calculation can be efficiently performed using numerical computing libraries (such as NumPy in Python). The results of eigenvalue decomposition provide a theoretical basis for the subsequent dimensionality reduction process because it reveals the main sources of variation in the data.

[0121] Then, the eigenvectors are sorted in descending order according to the size of the eigenvalue. The eigenvectors corresponding to larger eigenvalues ​​contain the most important variation information in the data. By calculating the cumulative contribution rate, that is, the proportion of the eigenvalues ​​to the total variance accumulated in order, the number of principal components that need to be retained can be determined. Usually, a threshold of 85% or 90% cumulative contribution rate is selected as the cutoff point. For example, if the cumulative contribution rate of the first three principal components reaches 90%, then only these three principal components can be retained and the remaining principal components can be discarded. The advantage of this step is that it provides an objective criterion to balance the trade-off between information retention and dimensionality reduction.

[0122] After determining the number of principal components to be retained, the next step is to use these principal components to perform a projection transformation on the standardized electrochemical characteristic indicators. This process actually projects the original m-dimensional data into a low-dimensional space spanned by the selected principal components. The specific operation is to multiply the original data matrix with the transformation matrix composed of the selected eigenvectors. The result is the principal component score after preliminary dimensionality reduction, and each sample has a new coordinate in the new principal component space. The advantage of this transformation is that it retains the most important variation information in the data, while significantly reducing the dimensionality of the data, which is conducive to subsequent analysis and visualization.

[0123] In order to further improve the interpretability of the principal components, the principal component scores after the initial dimensionality reduction are orthogonally rotated. Commonly used rotation methods include variance maximization rotation (Varimax) and oblique rotation (Promax). Taking Varimax rotation as an example, its goal is to make the load on each principal component present a highly differentiated distribution, that is, some variables have a high load on this principal component, while the load of other variables is close to zero. This rotation helps to identify the specific physical or chemical meaning represented by each principal component. For example, after rotation, it may be found that the first principal component is mainly related to surface activity, the second principal component reflects surface uniformity, and the third principal component may represent surface stability.

[0124] Finally, the rotated principal components are used as the electrochemical characteristic index set after dimensionality reduction. This new index set not only has a greatly reduced dimension, but also each principal component has a clear physical meaning. For example, for a data set that originally had 10 features, only 3 principal components may be retained in the end, representing surface activity, uniformity, and stability, respectively. This dimensionality reduction not only greatly reduces the computational complexity of subsequent analysis, but also provides a new perspective to understand the key influencing factors of copper alloy surface drawing quality.

[0125] This PCA-based data processing method has multiple advantages. First, it effectively solves the "curse of dimensionality" problem in high-dimensional data analysis, allowing subsequent machine learning algorithms to run more efficiently. Second, by retaining the most important variation information, this method maximizes the retention of key features of the original data while reducing the data dimension. Furthermore, the rotated principal components often have better interpretability, which helps researchers to deeply understand the key factors affecting the surface drawing quality of copper alloys. Finally, the output results of this method (i.e., the set of electrochemical characteristic indicators after dimensionality reduction) provide high-quality input features for subsequent quality classification and prediction models, which is expected to significantly improve the performance and generalization ability of the model. In general, this data processing method not only improves the efficiency and accuracy of the analysis, but also provides a more scientific and systematic framework for the evaluation of the surface drawing quality of copper alloys.

[0126] Please continue reading Figure 1 , the electrochemical characteristic index set is input into a support vector machine classification model pre-trained based on sample data of known quality grade, and a classification result representing the quality grade of the current sample is output.

[0127] In one embodiment of the present invention, the training process of the support vector machine classification model includes: performing electrochemical measurement and data processing on copper alloy drawing samples of known quality grade to obtain a training data set with quality grade labels; performing feature normalization and dimensionality reduction processing on the training data set to obtain a preprocessed feature matrix; constructing a kernel matrix based on the preprocessed feature matrix, and performing grid search on different kernel functions and parameter combinations to obtain the optimal kernel function configuration; using the optimal kernel function configuration and the preprocessed feature matrix, solving the dual problem of the support vector machine through a sequential minimum optimization algorithm to obtain support vectors and decision hyperplanes; and performing probability calibration on the support vectors and decision hyperplanes to obtain a support vector machine classification model with probability output capability.

[0128] Specifically, this step involves collecting a large number of samples with different quality levels (such as excellent, good, medium, and poor), and performing electrochemical measurements such as open circuit potential measurement, electrochemical impedance spectroscopy test, and cyclic voltammetry test on each sample. Subsequently, the electrochemical characteristic index set of each sample is obtained through the data processing methods described above, including parameter extraction, feature calculation, and principal component analysis. These characteristic index sets, together with the corresponding quality grade labels, constitute a training data set with quality grade labels. For example, a training data set containing 1,000 samples may consist of 10 electrochemical features and corresponding quality grade labels, forming a 1,000×11 matrix.

[0129] Next, the training data set is subjected to feature normalization and dimensionality reduction. Feature normalization is the process of converting features of different dimensions and scales to the same scale. Common methods include minimum-maximum normalization and Z-score normalization. This step ensures that different features have the same weight in model training. For example, the open circuit potential ranges from -0.5V to 0.5V and is normalized to the range of 0 to 1, while the charge transfer resistance ranges from 100Ω to 1000Ω and is also normalized to 0 to 1. Dimensionality reduction is to reduce the number of features and retain the most important information through methods such as principal component analysis. The combination of these two steps not only improves the efficiency of subsequent calculations, but also helps to reduce the impact of noise and redundant information, thereby obtaining a preprocessed feature matrix.

[0130] According to the preprocessed feature matrix, the next step is to construct the kernel matrix and select and optimize the kernel function. The kernel function is a key component in the SVM algorithm. It maps the original feature space to a high-dimensional space, making the originally linearly inseparable data linearly separable in the high-dimensional space. Commonly used kernel functions include linear kernels, polynomial kernels, and radial basis function (RBF) kernels. In order to find the kernel function and its parameters that best suit the current data, a grid search method is usually used. For example, for the RBF kernel, different γ values ​​(such as 0.1, 1, 10) and different C values ​​(such as 0.1, 1, 10, 100) may be tried, and the performance of each combination is evaluated through cross-validation, and finally the kernel function configuration with the best performance is selected. Although this systematic search method is computationally intensive, it can effectively avoid the bias caused by human selection and find the kernel function that is truly suitable for the data characteristics.

[0131] After obtaining the optimal kernel function configuration, this configuration and the preprocessed feature matrix are used to solve the dual problem of the support vector machine through the sequential minimum optimization (SMO) algorithm. The SMO algorithm is an efficient method for solving quadratic programming problems. It optimizes by iteratively selecting variable pairs that violate the KKT condition and eventually converges to the global optimal solution. The output of this process is a set of support vectors and corresponding weights, which together define the decision hyperplane. In the scenario of copper alloy surface drawing quality classification, the decision hyperplane can be understood as the boundary that separates different quality levels in the high-dimensional feature space. For example, in a two-classification problem (such as distinguishing high-quality and low-quality samples), the decision hyperplane is a high-dimensional plane that separates these two types of samples.

[0132] Finally, in order for the SVM model to output probabilistic prediction results rather than just category labels, the support vector and decision hyperplane need to be probabilistically calibrated. The commonly used method is Platt scaling, which estimates the category probability by fitting a sigmoid function on the output of the SVM decision function. The advantage of this step is that it makes the output of the model more interpretable and practical. For example, for a new sample, the model can not only tell which quality level it belongs to, but also provide the probability of belonging to that level (such as 80% probability of belonging to the high-quality category).

[0133] This SVM-based classification model training method has multiple advantages. First, SVM has good performance for high-dimensional data and is particularly suitable for processing complex multidimensional data such as electrochemical characteristics. Second, through the kernel technique, SVM is able to handle nonlinear classification problems, which is crucial for capturing the complex nonlinear relationship between the surface quality and electrochemical characteristics of copper alloys. Furthermore, the decision boundary of SVM is determined by the support vector, which makes the model insensitive to outliers and improves the robustness of classification. Finally, the introduction of probabilistic output increases the interpretability and practicality of the model prediction results, providing more information for quality control decisions. In general, this method can not only accurately classify the surface drawing quality of copper alloys, but also provide confidence in the classification results, providing strong support for quality control and process optimization in actual production.

[0134] In one embodiment of the present invention, the electrochemical characteristic indicator set is input into a support vector machine classification model pre-trained based on sample data of known quality grades, and a classification result representing the quality grade of the current sample is output, including: normalizing and reducing the dimension of the electrochemical characteristic indicator set to obtain a standardized feature vector consistent with the training data format; calculating the kernel matrix of the current sample and the support vector according to the standardized feature vector and the kernel function of the support vector machine classification model to obtain a decision function value; applying a probability calibration function to the decision function value to obtain a probability distribution of the sample belonging to each quality grade; performing multi-level threshold judgment according to the probability distribution to obtain the primary quality grade and secondary quality grade of the sample; combining the primary quality grade, secondary quality grade and their corresponding probability values ​​to obtain a classification result representing the quality grade of the current sample.

[0135] Specifically, the obtained electrochemical characteristic index set needs to be normalized and dimensionally reduced first. The purpose of this step is to ensure that the features of the current sample have the same scale and dimension as the training data, so as to ensure the accuracy of subsequent classification. Normalization usually uses the same method as the training data, such as minimum-maximum normalization or Z-score standardization. Assuming that the open circuit potential of a sample is -0.3V, and the minimum value of the open circuit potential in the training data is -0.5V and the maximum value is 0.5V, then after minimum-maximum normalization, the open circuit potential characteristic value of this sample will become 0.2. This treatment ensures that all features are in the same numerical range, avoiding some features from occupying inappropriate weights in the classification process due to large values. Dimensionality reduction usually uses the principal component analysis (PCA) method to project the original high-dimensional features into the principal component space determined when training the model. For example, if the original data has 20 features, and 5 principal components are determined to be retained during model training, then the 20 original features of the current sample will be converted into 5 principal component scores. This dimensionality reduction not only reduces the computational complexity, but also helps to eliminate noise and redundant information, and improve the stability and generalization ability of classification.

[0136] Next, based on the standardized feature vector and the kernel function of the support vector machine classification model, the kernel matrix of the current sample and the support vector is calculated to obtain the decision function value. This step actually maps the current sample to the high-dimensional feature space of the SVM model and calculates its distance from the decision hyperplane. Specifically, if the radial basis function (RBF) kernel is used, the kernel function value between the current sample and each support vector needs to be calculated in the form of ,in is the feature vector of the current sample, is the ith support vector, is the optimal parameter determined during model training. These kernel function values ​​are multiplied and summed with the corresponding Lagrange multiplier and category label to obtain the decision function value. The positive or negative value of the decision function value indicates which side of the decision hyperplane the sample is on, and its absolute value reflects the distance between the sample and the decision boundary, which provides a basis for subsequent probability estimation.

[0137] Then, a probability calibration function is applied to the decision function value to obtain the probability distribution of the sample belonging to each quality level. A commonly used probability calibration method is Platt scaling, which converts the decision value into a probability by fitting a sigmoid function on the SVM output. For example, for a binary classification problem (such as high quality vs. low quality), the sigmoid function may convert a decision function value of 0.8 into a probability of 90% belonging to the high quality category. In multi-classification problems, a "one-to-one" or "one-to-many" strategy is usually adopted to finally obtain the probability of the sample belonging to each quality level. This probability output provides richer information than simple category predictions, which helps in more detailed quality assessment and decision making.

[0138] Based on the obtained probability distribution, multi-level threshold judgment is performed to determine the primary quality level and secondary quality level of the sample. The introduction of this step reflects a more detailed understanding of the classification results. For example, it can be set that if the probability of a certain quality level exceeds 60%, it will be judged as the primary quality level; if the second highest probability exceeds 30%, it will be judged as the secondary quality level. The advantage of this multi-level judgment is that it can capture subtle differences and transition states in sample quality. For example, a sample may be judged as having a primary quality level of "good" (probability 65%) and a secondary quality level of "excellent" (probability 32%), which indicates that although the sample is of a good level overall, it is close to the excellent standard.

[0139] Finally, the primary quality grade, secondary quality grade and their corresponding probability values ​​are combined to obtain a complete classification result representing the quality grade of the current sample. This representation of the result is richer and more informative than a single category label. For example, a complete classification result may be expressed as: "Primary quality grade: good (65%), secondary quality grade: excellent (32%)". This detailed output provides quality control personnel with more comprehensive information, which helps to make more accurate judgments and decisions.

[0140] This support vector machine-based classification method combines the advancement of machine learning with the practicality of traditional quality control. It can not only accurately classify the quality of copper alloy surface drawing, but also provide confidence and secondary trends of the classification results, providing strong support for quality control and process optimization in actual production. In addition, this method is also very flexible, and the thresholds and judgment criteria can be adjusted according to specific production needs to adapt to different quality control strategies. In general, this comprehensive quality assessment method provides a scientific, reliable and insightful quality control tool for copper alloy surface drawing production, which is expected to significantly improve the consistency and predictability of product quality.

[0141] In one embodiment of the present invention, performing multi-level threshold judgment according to the probability distribution to obtain the primary quality grade and secondary quality grade of the sample includes: performing entropy analysis on the probability distribution to obtain an uncertainty index of the sample quality grade distribution; selecting a dynamic threshold strategy according to the uncertainty index, using a single threshold method for low uncertainty samples and a multiple threshold method for high uncertainty samples to obtain a threshold judgment scheme; applying the threshold judgment scheme to the probability distribution to obtain a preliminary quality grade judgment result; performing a neighborhood consistency check on the preliminary quality grade judgment result, performing local smoothing according to the quality grade distribution of adjacent measurement positions to obtain a spatially consistent quality grade distribution; extracting the grade with the highest probability from the spatially consistent quality grade distribution as the primary quality grade, and extracting the grade with the second highest probability and exceeding a preset threshold as the secondary quality grade to obtain the primary quality grade and secondary quality grade of the sample.

[0142] Specifically, first, entropy analysis is performed on the probability distribution, and information entropy is calculated as an uncertainty index for the sample quality grade distribution. Information entropy is a measure of the uncertainty of a random variable, and is used here to quantify the degree of ambiguity in quality grade judgment. For example, if the quality grade probability distribution of a sample is [0.9, 0.05, 0.03, 0.02] (corresponding to the four grades of excellent, good, medium, and poor), then the entropy value of this distribution will be lower, indicating a higher certainty in the judgment; on the contrary, if the probability distribution is [0.3, 0.25, 0.23, 0.22], then the entropy value will be higher, indicating a greater uncertainty in the judgment. The advantage of this entropy analysis is that it provides a quantitative indicator to evaluate the credibility of the classification results, providing an important basis for subsequent decision-making.

[0143] According to the calculated uncertainty index, the dynamic threshold strategy is selected next. The core idea of ​​this strategy is to adopt different judgment methods according to the degree of certainty of sample classification. For low-uncertainty samples, that is, samples with low entropy values, a single threshold method is used. For example, the level with the highest probability can be simply used as the final judgment result. For high-uncertainty samples, a multiple threshold method is used. This method considers multiple probability values ​​and may give multiple possible quality levels. For example, for a high-uncertainty sample, a judgment result such as "70% probability is good, 25% probability is excellent" may be given. The advantage of this dynamic threshold strategy is that it can flexibly adjust the judgment criteria according to the specific circumstances of the sample, which not only ensures the rapid judgment of clear samples, but also gives a more cautious and comprehensive evaluation of ambiguous samples.

[0144] Apply the determined threshold judgment scheme to the probability distribution to obtain the preliminary quality grade judgment result. This step converts the previous theoretical analysis into a specific classification result. For example, for a low uncertainty sample, if the highest probability exceeds the preset threshold (such as 80%), it will be directly judged as this grade; while for a high uncertainty sample, multiple grades with higher probabilities may be retained to provide more information for subsequent analysis. The advantage of this method is that it takes into account the main trend of the probability distribution and retains the secondary information in the distribution, providing a more comprehensive judgment basis for quality control.

[0145] Next, the preliminary quality grade judgment results are subjected to a neighborhood consistency check. This step takes into account the spatial continuity characteristics of the sample surface quality. Specifically, for the judgment result of each measurement position, the quality grade distribution of its adjacent positions must be considered. For example, if a position is judged as "good" and most of the positions around it are judged as "excellent", then this "good" judgment may need to be re-evaluated. By locally smoothing the quality grade distribution of adjacent measurement positions, a spatially consistent quality grade distribution can be obtained. The advantage of this neighborhood consistency check is that it can eliminate local abnormal judgments and improve the reliability and consistency of the overall judgment results. At the same time, this method also helps to identify the quality change trend of the sample surface and provide valuable information for process improvement.

[0146] Finally, the final judgment result is extracted from the spatially consistent quality grade distribution. Specifically, the grade with the highest probability is determined as the main quality grade, and the grade with the second highest probability and exceeding the preset threshold is extracted as the secondary quality grade. For example, if the probability distribution of a sample after spatial consistency processing is [0.65, 0.30, 0.03, 0.02], and the threshold of the secondary grade is set to 25%, then the main quality grade of this sample will be judged as "excellent" and the secondary quality grade will be "good". This dual-grade judgment method provides a more detailed and comprehensive quality assessment result. It not only points out the main quality level of the sample, but also reflects the potential change trend or transition state of the sample quality, providing richer information for quality control and process optimization.

[0147] This comprehensive quality grade judgment method has multiple advantages. First, it introduces uncertainty assessment through entropy analysis, which improves the reliability of judgment. Second, the application of dynamic threshold strategy enables the judgment process to flexibly adapt to different situations, ensuring both efficiency and accuracy. Third, the neighborhood consistency check takes into account the spatial correlation of sample surface quality, which helps to obtain more robust and coherent judgment results. Finally, the dual-grade output provides more detailed and comprehensive information for quality control. In general, this method can not only accurately evaluate the quality of copper alloy surface drawing, but also provide the spatial characteristics and potential change trends of quality distribution, providing strong support for the optimization of production processes and quality control.

[0148] Another embodiment of the present application further provides a copper alloy surface wire drawing quality detection device, which is used to implement the copper alloy surface wire drawing quality detection method provided in the above embodiment.

[0149] Figure 2 The present invention is a schematic diagram of a structure of a copper alloy surface drawing quality detection device; specifically, the copper alloy surface drawing quality detection device comprises:

[0150] The sample pretreatment unit 101 is used to perform surface cleaning and drying treatment on the copper alloy wire drawing sample to obtain a clean sample;

[0151] An electrochemical measurement module 102 is used to select a plurality of predetermined positions on the surface of the clean sample based on surface feature analysis, perform electrochemical measurement on each predetermined position in turn, and obtain a plurality of sets of electrochemical measurement data, wherein the electrochemical measurement includes open circuit potential measurement, electrochemical impedance spectroscopy test and cyclic voltammetry test;

[0152] The data processing unit 103 is used to perform parameter extraction and mathematical operations on the multiple sets of electrochemical measurement data to obtain a set of electrochemical characteristic indicators that characterize the surface characteristics of the current sample;

[0153] The quality classification module 104 is used to input the electrochemical characteristic index set into a support vector machine classification model pre-trained based on sample data of known quality grades, and output a classification result representing the quality grade of the current sample.

[0154] Specifically, in the copper alloy surface drawing quality inspection system, the sample pretreatment unit 101 first performs preliminary treatment on the copper alloy drawing sample to ensure that all samples to be tested are in the same initial state. The unit is equipped with an ultrasonic cleaner and a vacuum drying oven, which effectively removes pollutants, oil stains and moisture on the sample surface through anhydrous ethanol cleaning and high-purity nitrogen purging. In addition, the sample pretreatment unit 101 also includes a precision cutting device for adjusting the sample to a size suitable for measurement. After these treatments, the sample is transferred to the electrochemical measurement module 102, which is the core of the entire system.

[0155] The electrochemical measurement module 102 is equipped with a movable micro-electrolytic cell and a high-precision electrochemical workstation. The micro-electrolytic cell is accurately placed at a predetermined position on the sample surface through a precision positioning system, while the electrochemical workstation is responsible for performing open circuit potential measurement, electrochemical impedance spectroscopy test and cyclic voltammetry test. This module also contains a multi-channel data acquisition system that can simultaneously record multiple electrochemical parameters such as potential, current and impedance. The design of the electrochemical measurement module 102 fully considers the accuracy and repeatability of the measurement, and ensures that the conditions of each measurement remain consistent through an automated control system.

[0156] After the measurement is completed, the raw data is transferred to the data processing unit 103 for in-depth analysis. The data processing unit 103 is equipped with a high-performance computing server running specially developed data processing software. The software first pre-processes the raw data, including denoising and outlier detection. Then, key electrochemical features such as charge transfer resistance and double layer capacitance are extracted from the processed data through complex algorithms. Subsequently, these features are standardized to ensure that features of different scales can be effectively compared. Finally, the data processing unit 103 uses dimensionality reduction techniques such as principal component analysis to convert the high-dimensional data into a more compact feature representation.

[0157] The processed data is finally input into the quality classification module 104, which is the decision-making unit of the system. The quality classification module 104 integrates a pre-trained support vector machine model, which is trained by a large amount of sample data with known quality grades. The module first calculates the probability of the sample belonging to each quality grade, and then performs multi-level threshold judgment based on these probabilities to determine the primary and secondary quality grades of the sample. The quality classification module 104 is also equipped with a result visualization system that can generate intuitive quality assessment reports, including quality grade distribution diagrams and key parameter analysis.

[0158] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. All equivalent structural changes made by using the contents of the present invention specification and drawings under the inventive concept of the present invention, or directly / indirectly applied in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. A copper alloy surface drawing quality detection method, characterized in that: include: The copper alloy wire drawing samples were surface cleaned and dried to obtain clean samples; Based on the surface feature analysis, multiple predetermined positions are selected on the surface of the clean sample, and electrochemical measurements are performed on each predetermined position in turn to obtain multiple groups of electrochemical measurement data, wherein the electrochemical measurement includes open circuit potential measurement, electrochemical impedance spectrum test and cyclic voltammetry test, and the multiple predetermined positions are selected on the surface of the clean sample based on the surface feature analysis, including: performing three-dimensional laser scanning on the surface of the clean sample to obtain a microscopic topographic map; performing curvature analysis and shape analysis on the surface of the clean sample according to the microscopic topographic map to obtain a surface feature distribution map; performing grid division and spatial clustering based on the surface feature distribution map to obtain a candidate measurement point set; performing Voronoi diagram analysis and spatial autocorrelation analysis on the candidate measurement point set to obtain an optimized measurement point set; determining the points in the optimized measurement point set as predetermined positions to obtain multiple predetermined positions; The electrochemical measurement is performed on each predetermined position in sequence to obtain multiple sets of electrochemical measurement data, including: setting a movable electrolytic cell at the current predetermined position, using the sample surface exposed in the electrolytic cell as a working electrode, placing a reference electrode and a counter electrode in the electrolytic cell, and injecting an electrolyte into the electrolytic cell to form a local three-electrode measurement system; applying alternating potential and current scanning to the working electrode in the local three-electrode measurement system to obtain a working electrode with stable electrochemical activity; sequentially performing open circuit potential measurement, electrochemical impedance spectrum test and cyclic voltammetry test on the working electrode with stable electrochemical activity to obtain a set of electrochemical measurement data for the current predetermined position; moving the movable electrolytic cell to the next predetermined position, repeating the above steps of electrolytic cell setting, electrode activation and electrochemical measurement until the measurement of all predetermined positions is completed; summarizing the electrochemical measurement data of all predetermined positions in position order to obtain multiple sets of electrochemical measurement data including open circuit potential data, impedance spectrum data and voltammetric curve data corresponding to each predetermined position; Performing parameter extraction and mathematical operations on the multiple sets of electrochemical measurement data to obtain a set of electrochemical characteristic indicators characterizing the surface characteristics of the current sample; The electrochemical characteristic index set is input into a support vector machine classification model pre-trained based on sample data of known quality grades, and a classification result representing the quality grade of the current sample is output.

2. The copper alloy surface drawing quality detection method according to claim 1, characterized in that: The electrochemical impedance spectroscopy test includes: Applying multi-frequency AC potential perturbations to the working electrode with stable electrochemical activity to obtain preliminary impedance response data in a wide frequency domain; Calculate the frequency derivative of impedance based on the broadband preliminary impedance response data, identify the characteristic frequency interval where impedance changes significantly, and obtain the key frequency range; Perform high-resolution frequency scanning within the critical frequency range to obtain local fine impedance response data; The broadband preliminary impedance response data and the local fine impedance response data are combined to construct a composite impedance spectrum to obtain impedance spectrum data.

3. The copper alloy surface drawing quality detection method according to claim 1, characterized in that: The said performing parameter extraction and mathematical operation on the said plurality of sets of electrochemical measurement data to obtain a set of electrochemical characteristic indicators characterizing the surface characteristics of the current sample includes: Performing statistical analysis on the open circuit potential data of the plurality of groups of electrochemical measurement data, extracting the potential mean and fluctuation range, and obtaining a surface electrochemical stability index; Performing equivalent circuit fitting on the impedance spectrum data of the multiple sets of electrochemical measurement data, extracting charge transfer resistance and double layer capacitance values, and obtaining interface electrochemical characteristic indicators; Performing peak analysis and integral calculation on the voltammetric curve data of the multiple sets of electrochemical measurement data, extracting redox peak potential, peak current density and reaction charge, and obtaining electrochemical activity indicators; Normalizing the surface electrochemical stability index, the interface electrochemical characteristic index and the electrochemical activity index to obtain a standardized electrochemical characteristic index; The standardized electrochemical characteristic indexes are subjected to principal component analysis to extract main characteristic vectors, thereby obtaining a set of electrochemical characteristic indexes after dimensionality reduction.

4. The copper alloy surface drawing quality detection method according to claim 3, characterized in that: The principal component analysis is performed on the standardized electrochemical characteristic index to extract the main characteristic vectors to obtain a set of electrochemical characteristic indicators after dimension reduction, including: Constructing a covariance matrix of the standardized electrochemical characteristic index, performing eigenvalue decomposition, and obtaining eigenvalues ​​and corresponding eigenvectors; Sort the eigenvectors according to the eigenvalues, calculate the cumulative contribution rate, and determine the number of principal components to be retained; Performing a projection transformation on the standardized electrochemical characteristic index using the determined number of principal components to obtain a principal component score after preliminary dimensionality reduction; Performing orthogonal rotation on the principal component scores after the preliminary dimensionality reduction to obtain rotated principal components; The rotated principal components are used as a set of electrochemical characteristic indicators after dimensionality reduction.

5. The copper alloy surface drawing quality detection method according to claim 1, characterized in that: The training process of the support vector machine classification model includes: Conduct electrochemical measurements and data processing on copper alloy wire drawing samples with known quality grades to obtain a training data set with quality grade labels. Performing feature normalization and dimensionality reduction processing on the training data set to obtain a preprocessed feature matrix; Constructing a kernel matrix according to the preprocessed feature matrix, performing grid search on different kernel functions and parameter combinations to obtain an optimal kernel function configuration; Using the optimal kernel function configuration and the preprocessed feature matrix, the dual problem of the support vector machine is solved by a sequential minimum optimization algorithm to obtain a support vector and a decision hyperplane; The support vector and the decision hyperplane are probability calibrated to obtain a support vector machine classification model with probability output capability.

6. The copper alloy surface drawing quality detection method according to claim 1, characterized in that: The electrochemical characteristic index set is input into a support vector machine classification model pre-trained based on sample data of known quality grade, and a classification result indicating the quality grade of the current sample is output, including: Normalizing and reducing the dimension of the electrochemical characteristic index set to obtain a standardized feature vector consistent with the training data format; According to the standardized feature vector and the kernel function of the support vector machine classification model, a kernel matrix of the current sample and the support vector is calculated to obtain a decision function value; Applying a probability calibration function to the decision function value to obtain a probability distribution of the sample belonging to each quality grade; Performing multi-level threshold judgment according to the probability distribution to obtain the primary quality grade and secondary quality grade of the sample; The primary quality grade, the secondary quality grade and their corresponding probability values ​​are combined to obtain a classification result representing the quality grade of the current sample.

7. The copper alloy surface drawing quality detection method according to claim 6, characterized in that: The performing of multi-level threshold judgment according to the probability distribution to obtain the primary quality grade and the secondary quality grade of the sample includes: Performing entropy analysis on the probability distribution to obtain an uncertainty index of the sample quality grade distribution; A dynamic threshold strategy is selected according to the uncertainty index, a single threshold method is used for low uncertainty samples, and a multiple threshold method is used for high uncertainty samples to obtain a threshold judgment scheme; Applying the threshold judgment scheme to the probability distribution to obtain a preliminary quality grade judgment result; Performing a neighborhood consistency check on the preliminary quality grade judgment result, and performing local smoothing according to the quality grade distribution of adjacent measurement positions to obtain a spatially consistent quality grade distribution; The grade with the highest probability is extracted from the spatially consistent quality grade distribution as the primary quality grade, and the grade with the second highest probability and exceeding a preset threshold is extracted as the secondary quality grade, to obtain the primary quality grade and secondary quality grade of the sample.

8. A copper alloy surface drawing quality detection device, characterized in that: The copper alloy surface drawing quality detection device is applied to the copper alloy surface drawing quality detection method according to any one of claims 1 to 7, comprising: The sample pretreatment unit is used to clean and dry the surface of the copper alloy wire drawing sample to obtain a clean sample; An electrochemical measurement module, for selecting a plurality of predetermined positions on the surface of the clean sample based on surface feature analysis, performing electrochemical measurement on each predetermined position in turn, and obtaining a plurality of sets of electrochemical measurement data, wherein the electrochemical measurement includes open circuit potential measurement, electrochemical impedance spectroscopy test and cyclic voltammetry test; A data processing unit, used to perform parameter extraction and mathematical operations on the multiple sets of electrochemical measurement data to obtain a set of electrochemical characteristic indicators characterizing the surface characteristics of the current sample; The quality classification module is used to input the electrochemical characteristic index set into a support vector machine classification model pre-trained based on sample data of known quality grades, and output a classification result representing the quality grade of the current sample.

Images